NVIDIA DGX Station GB300

Brand: NVIDIA | Category: Workstations

SKU: NVID-900216000000000 | Part #: 900-21600-0000-000 | MPN: 900-21600-0000-000

Contact for Pricing — Request a Quote

Request a Quote Contact Us

About the NVIDIA DGX Station GB300

The NVIDIA DGX Station GB300 from NVIDIA is enterprise-grade Workstations hardware built for data centers. Grace Blackwell Ultra desktop AI supercomputer delivering 784 GB of coherent unified memory for large-scale model training and inference at the workstation. Available in brand new condition. Available from Omnixon Global — contact our team to request a quote.

Technical Specifications

BrandNVIDIA
CategoryWorkstations
SKUNVID-900216000000000
Part Number900-21600-0000-000
ConditionNew
Product LineDGX Station
GenerationGB300 Grace Blackwell Ultra
Form FactorDesktop workstation (tower)
GPUNVIDIA GB300 Blackwell Ultra GPU
CPUNVIDIA Grace Arm-based CPU (integrated via NVLink-C2C)
CPU-GPU InterconnectNVLink-C2C (chip-to-chip coherent interconnect)
Unified Memory Capacity784 GB coherent HBM + LPDDR5X unified memory pool
GPU Memory TypeHBM3e
Memory BandwidthMulti-TB/s aggregate NVLink-C2C + HBM bandwidth
FP8 Tensor Core PerformanceBlackwell Ultra 5th-generation Tensor Cores with FP8, FP16, BF16, TF32 support
System StorageMultiple high-capacity NVMe SSDs (RAID configuration)
NetworkingHigh-speed Ethernet and/or InfiniBand (ConnectX-series NIC)
PCIe GenerationPCIe Gen 5
Operating SystemDGX OS (Ubuntu-based Linux)
Software StackNVIDIA AI Enterprise, NeMo, TensorRT-LLM, cuDNN, CUDA, RAPIDS
Container RegistryNVIDIA NGC container support included
Power SupplyHigh-efficiency multi-rail PSU designed for sustained AI compute workloads
CoolingActive managed cooling with precision airflow engineering for desktop operation
ManagementNVIDIA BMC with IPMI, Redfish API, and DGX systems management tools
Target WorkloadsLLM training and fine-tuning, multimodal inference, genomics, digital twins, agentic AI
Announced2025-Q4